Topaz: Image Enhance Bloom 2
Enhance AI-generated images with Bloom 2, a creative upscaling model that adds new detail, texture, and visual elements while offering a Creativity control from 1–9 for subtle refinement or bold reinterpretation. It accepts 1 input image and produces 1 enhanced output, with optional Face Recovery, Color Preservation, and grain settings for fine-tuned results. Ideal for improving AI artwork and illustrations, adding realism to CG renders, and refining facial detail in generated portraits.
Try an example
About this workflow
This workflow enhances a single input image using the TopazImageEnhanceV2 node configured with the Bloom 2 model. Bloom 2 is a creative enhancement and upscaling model that can add plausible detail, texture, and visual elements to AI art, illustrations, portraits, and CG renders. The key control is the Creativity setting (1–9), which lets you choose between subtle refinement and bold reinterpretation. Optional toggles for Face Recovery and Color Preservation help maintain natural facial features and original color palettes, while a Grain control can add film-like texture to reduce a too-clean, synthetic look.
Technically, the pipeline is straightforward: LoadImage brings your source into the graph, TopazImageEnhanceV2 (set to Bloom 2) performs the enhancement, ImageCompare previews the before/after for quick A/B evaluation, and SaveImageAdvanced writes the result with custom naming and format options. A MarkdownNote in the workflow provides quick tips, including the note that if the prompt field is left empty, the model will auto-generate guidance. This makes the workflow fast to adopt while still offering fine-grained control when you need it.
Questions & answers
Creativity (1–9) controls how much new detail and reinterpretation Bloom 2 adds. Use 1–3 for gentle cleanup, 4–6 for balanced, and 7–9 for stronger stylistic changes. Start low and increase gradually while checking ImageCompare to avoid over-processing.
Enable Preserve Color in the TopazImageEnhanceV2 node. This keeps hue and overall palette closer to your input. If the image still looks off, try lowering Creativity and disabling Grain, then reintroduce them incrementally.
Turn on Face Recovery and use a lower Creativity (around 2–4) for portraits. This combination restores facial structure while avoiding excessive hallucinated detail. Use ImageCompare to ensure expressions and identity remain consistent.
No. If you leave the prompt field empty, the model will auto-create guidance. If you want to steer the result, add a concise prompt (e.g., “clean linework, fine fabric texture, realistic skin pores”) to nudge the enhancement toward your target style.